The contribution of structured interview surveys to the monitoring of changes in risk behaviour is a crucial issue for the evaluation of HIV control programmes. A review is made of studies that have attempted to assess the reliability and validity of self-reported sexual behaviour. Evidence from developed-country studies is encouraging; in general the quality of responses appears to be as high as that found in studies of other topics. In developing countries, there have been fewer studies and results are more varied. One clear lesson is that survey execution must be of a very high standard; as in all survey research, but particularly in the case of sex surveys, poor standards of design and execution will yield untrustworthy data.
PIP: In the current context of a global HIV/AIDS pandemic, the authors consider whether large-scale surveys can make the same contribution to HIV control programs as they have to family planning programs. It is, however, very difficult to give clear-cut conclusions on the validity of survey information on sexual behavior and related matters. Only very limited, variable evidence exists to date, while no clear guidelines exist with regard to the nature and extent of error which can be tolerated. Nonetheless, the authors make a few general observations. First, only very few field or other checks of data quality have revealed error of a magnitude sufficient to suggest that the survey approach be abandoned. Second, the quality of data from sex surveys depends critically upon standards of survey execution; poorly designed and conducted surveys will always yield poor results. Survey research on risk behaviors related to HIV/STD transmission will never be a cheap option and the undertaking of low-cost surveys should not be encouraged. Furthermore, experimentation with new ways to measure sexual behaviors should continue. The WHO/GPA approach provides far too little detailed information on the many different types of sexual partnerships. The attempt to distinguish between contacts with sex workers and other contacts is also ill-developed. The authors close in noting the enormous potential for advances in survey design and their hope that funding agencies will intensify their support of such methodological work.